Year: 2016
Type: conference-paper
Abstract: In recent years, neural machine translation (NMT) has demonstrated state-of-the-art machine\ntranslation (MT) performance. It is a new approach to MT, which tries to learn a set of parameters\nto maximize the conditional probability of target sentences given source sentences. In this paper,\nwe present a novel approach to improve the translation performance in NMT by conveying topic\nknowledge during translation. The proposed topic-informed NMT can increase the likelihood of\nselecting words from the same topic and domain for translation. Experimentally, we demonstrate\nthat topic-informed NMT can achieve a 1.15 (3.3% relative) and 1.67 (5.4% relative) absolute\nimprovement in BLEU score on the Chinese-to-English language pair using NIST 2004 and 2005\ntest sets, respectively, compared to NMT without topic information.
Authors: Jian Zhang, Liangyou Li, Andy Way, Qun Liu
Institution: Dublin City University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2573728411